Ordered Polymorphisms: Theory & Applications

• Ordered polymorphisms are specialized polymorphisms that preserve an order relation or structured ordering, appearing in algebraic CSPs, Boolean PCSPs, and crystallographic studies.
• They enable tractable computational frameworks by enforcing monotonicity in CSPs and underpin algorithmic dichotomies through threshold and measure-preserving functions.
• In materials science and polymers, ordered polymorphism denotes the coexistence of multiple ordered structures or sequence morphologies, aiding in crystal structure prediction and design.

Searching arXiv for relevant papers on ordered polymorphisms and closely related notions. {"query":"all:(\"ordered polymorphisms\" OR polymorphisms CSP monotone polymorphisms Promise CSP polymorphism crystal structure prediction polymorphism infinite-dimensional groups)", "max_results": 10} {"query":"ordered polymorphisms", "max_results": 10} Ordered polymorphisms are not a single discipline-independent formal object. In the algebraic theory of CSPs and Promise CSPs, they usually mean polymorphisms that additionally preserve a designated order, most explicitly monotone operations preserving $\le$ on a poset; in Boolean Ordered PCSPs, the inclusion of the predicate $x \le y$ forces all polymorphisms to be monotone (Brady, 2022, Brakensiek et al., 2021). In adjacent literatures, the same phrase acquires different organizational meanings: polymorphisms can form semigroups and categories of measure-theoretic morphisms for nonsingular actions of large groups (Neretin, 2023), while materials papers use ordered polymorphism for multiple ordered crystal structures with the same composition and distinct space groups or symmetry lineages (Omee et al., 12 Jun 2025, Yamazaki et al., 23 Apr 2026). The term is therefore best interpreted relative to its formal context.

1. Terminological scope and principal uses

Several of the cited works explicitly state that they do not introduce a separate universal notion called “ordered polymorphism.” Instead, they supply setting-specific meanings. In the algebraic CSP literature, the closest direct notion is that ordered polymorphisms are polymorphisms that preserve an order relation, such as monotone operations on $(\{0,1\},\le)$ (Brady, 2022). In Boolean Ordered PCSPs, the “ordered” modifier is attached to the presence of the predicate $x \le y$, and the resulting polymorphisms are monotone Boolean functions (Brakensiek et al., 2021). In the measure-theoretic setting, the relevant content is organizational rather than order-theoretic: polymorphisms are arranged by topology, closure, semigroup composition, and category structure, but there is no lattice-theoretic or total-order notion on polymorphisms themselves (Neretin, 2023). In materials science, ordered polymorphism refers to recovering or classifying several ordered crystal structures of one composition rather than a single ground state (Omee et al., 12 Jun 2025).

Domain	What “ordered” means	Representative source
Algebraic CSP	Preserves $\le$; monotone polymorphism	(Brady, 2022)
Boolean Ordered PCSP	Predicate $x \le y$ forces monotone polymorphisms	(Brakensiek et al., 2021)
Promise CSP with ordered colours	Preserves the linearly ordered predicate $\mathrm{LO}_k$	(Krokhin et al., 20 Aug 2025)
Measure-theoretic dynamics	Organized by semigroup/category and closure	(Neretin, 2023)
Crystal structure prediction	Multiple ordered crystal structures of one composition	(Omee et al., 12 Jun 2025)
Order-(dis)order materials lineage	Ordered children of disordered parents; group-subgroup families	(Yamazaki et al., 23 Apr 2026)

A common misconception is that “ordered polymorphism” always denotes an order-preserving operation in the algebraic sense. The literature does not support that universal reading. A second misconception is the reverse one: that the phrase is purely metaphorical. In CSP and PCSP theory, the order relation can be literal and formally preserved; in materials science, the ordering is crystallographic and tied to symmetry, space groups, and family relations.

2. Order preservation in algebraic CSP theory

The foundational algebraic definition is relation preservation. A $k$-ary function $f$ preserves an $m$-ary relation $x \le y$0, written $x \le y$1, if for every choice of $x \le y$2 $x \le y$3-tuples in $x \le y$4, applying $x \le y$5 componentwise produces a new $x \le y$6-tuple that is also in $x \le y$7. A function is a polymorphism of $x \le y$8 if it preserves every relation in $x \le y$9 (Brady, 2022). This yields the standard Inv–Pol correspondence,

$(\{0,1\},\le)$0

together with the statement that there is an order reversing bijection between clones and relational clones, given by $(\{0,1\},\le)$1 and $(\{0,1\},\le)$2 (Brady, 2022).

Within this framework, the most direct algebraic realization of an ordered polymorphism is a monotone operation. The notes define an operation $(\{0,1\},\le)$3 to be monotone if it preserves the relation $(\{0,1\},\le)$4, and self-dual if it preserves the relation $(\{0,1\},\le)$5 (Brady, 2022). This makes monotone operations exactly the order-preserving polymorphisms of the poset $(\{0,1\},\le)$6. The same source proves that if a function is monotone and self-dual, then it lies in the majority clone, and also that if a relation is preserved by the majority function $(\{0,1\},\le)$7, then it is bijunctive (Brady, 2022).

The broader significance is the standard algebraic principle that CSP complexity is controlled by polymorphisms. The classic tractable cases listed in the notes are HORN-SAT with semilattice polymorphism $(\{0,1\},\le)$8, 2SAT with majority polymorphism $(\{0,1\},\le)$9, and XOR-SAT with affine/Mal’cev polymorphism $x \le y$0 (Brady, 2022). Ordered polymorphisms fit naturally into this landscape because preserving a poset relation is one specific preservation condition among the many identities that govern tractability. The same notes place monotone/order-preserving operations alongside near-unanimity, Mal’cev, edge, Taylor, cyclic, and height-1 identities, thereby situating ordered polymorphisms within the larger clone-theoretic and minion-theoretic apparatus (Brady, 2022).

This suggests a useful terminological distinction. In the strict algebraic sense, ordered polymorphisms are best treated as a special case of relation-preserving operations, where the distinguished relation is an order or preorder. In that sense they are not an alternative to the standard polymorphism formalism; they are an instance of it.

3. Ordered polymorphisms in Promise CSPs

Boolean Ordered PCSPs make the order relation explicit. The special feature is the presence of the predicate

$x \le y$1

which is exactly $x \le y$2. Preservation of this relation implies

$x \le y$3

so every polymorphism is a monotone Boolean function (Brakensiek et al., 2021). The paper proves that Boolean Ordered PCSPs exhibit a computational dichotomy assuming the Rich 2-to-1 Conjecture: such a PCSP is in polynomial time if for every $x \le y$4 it has polymorphisms where each coordinate has Shapley value at most $x \le y$5, and otherwise it is NP-hard (Brakensiek et al., 2021). An equivalent formulation given there is that polynomial-time solvability is equivalent to having threshold polymorphisms of arbitrarily large arity (Brakensiek et al., 2021).

The analytic mechanism is built from the Shapley value

$x \le y$6

together with a structural lemma asserting that Boolean monotone functions with each coordinate having low Shapley value have arbitrarily large threshold functions as minors (Brakensiek et al., 2021). The algorithmic side then proceeds through threshold polymorphisms and the known BLP+Affine machinery; the hardness side uses the consistency of Shapley value under a uniformly random 2-to-1 minor (Brakensiek et al., 2021). The conditional nature of this dichotomy is important: the paper states it under the Rich 2-to-1 Conjecture, not as an unconditional theorem.

A distinct ordered PCSP appears in linearly ordered hypergraph colouring. There the relevant predicate is

$x \le y$7

For $x \le y$8, this is exactly monotone 1-in-3 SAT: $x \le y$9 A polymorphism of $\le$0 is a function $\le$1 such that for every partition $\le$2 of $\le$3,

$\le$4

The paper proves that $\le$5 is NP-hard and derives as an immediate corollary that $\le$6 is NP-hard (Krokhin et al., 20 Aug 2025). Its structural analysis shows that, after excluding small 2-sets, polymorphisms behave like recoloured projections with a unique dictating variable, which is precisely the kind of rigidity that drives the hardness theorem (Krokhin et al., 20 Aug 2025).

These two PCSP developments show that “ordered polymorphisms” can mean either monotonicity with respect to the Boolean order or preservation of a richer ordered target predicate. In both cases, the order constraint sharply restricts the minion of admissible operations.

4. Fractional and approximate polymorphisms in optimization

In combinatorial optimization, the relevant objects are fractional and approximate polymorphisms rather than order-preserving operations. A $\le$7-ary polymorphism is an operation

$\le$8

such that coordinatewise application to satisfying assignments of a CSP instance yields another satisfying assignment (Brown-Cohen et al., 2015). For optimization, the paper introduces approximate polymorphisms and fractional polymorphisms through the inequality

$\le$9

capturing the idea that the operation “never increases the value” or “does not increase the expected value” of $x \le y$0 (Brown-Cohen et al., 2015).

The main structural conditions are not order-theoretic but measure-preserving and transitive symmetry. Measure-preserving means that for each symbol $x \le y$1, the fraction of inputs equal to $x \le y$2 equals the fraction of outputs equal to $x \le y$3, for every choice of inputs. Transitive symmetry means that for all positions $x \le y$4, there is a permutation $x \le y$5 with $x \le y$6 and

$x \le y$7

The main theorem states: if $x \le y$8 admits a fractional polymorphism whose support consists of measure-preserving, transitive symmetric operations, then there is an efficient randomized algorithm making $x \le y$9 value-oracle queries and outputting $\mathrm{LO}_k$0 such that

$\mathrm{LO}_k$1

(Brown-Cohen et al., 2015).

The same paper states that it does not introduce a notion called ordered polymorphism. If one looks for the nearest analogue, the closest concept is transitive symmetry, which is coordinate-permutation invariance rather than preservation of an external order relation (Brown-Cohen et al., 2015). A plausible implication is that, in optimization, “structure on polymorphisms” is often conveyed through symmetry constraints and marginal constraints rather than through monotonicity.

5. Measure-theoretic polymorphisms and categorical organization

In ergodic-theoretic and representation-theoretic work, a polymorphism is a measure on

$\mathrm{LO}_k$2

satisfying the pushforward conditions

$\mathrm{LO}_k$3

(Neretin, 2023). Specializing to a single space $\mathrm{LO}_k$4 gives the formulation

$\mathrm{LO}_k$5

These objects are “spread-out maps” carrying Radon–Nikodym weights; a measure-preserving polymorphism is supported on $\mathrm{LO}_k$6 (Neretin, 2023).

The decisive structure is semigroup and category structure. If

$\mathrm{LO}_k$7

then their product

$\mathrm{LO}_k$8

is defined by a convolution-type composition. For continuous polymorphisms, if $\mathrm{LO}_k$9 and $k$0 are determined by measurable families of measures $k$1 and $k$2 on $k$3, then

$k$4

where $k$5 is convolution of measures on $k$6 (Neretin, 2023). The paper states that multiplication extends by separate continuity to all polymorphisms, is associative, and hence polymorphisms form a category whose objects are Lebesgue probability spaces and whose morphisms are polymorphisms (Neretin, 2023).

This paper is explicit that there is no separate order-theoretic notion of ordered polymorphism. The relevant organizing content comes instead from topology and closure, from semigroup composition, and from the train $k$7 of reduced double cosets $k$8 (Neretin, 2023). Its main theorem states that if $k$9 acts nonsingularly on $f$0 and the subgroup $f$1 acts measure-preservingly, then

$f$2

defines a functor from the train to the category of polymorphisms (Neretin, 2023). In this literature, therefore, ordered polymorphism is best understood, if at all, in the broad sense of structured organization rather than as a preserved partial order.

6. Ordered polymorphism in materials and polymer science

In crystal structure prediction, ordered polymorphism means recovering multiple ordered crystal structures of one composition rather than treating every structure except the global minimum as wrong. The paper on ParetoCSP2 defines the task in practice as predicting multiple candidate structures that share the same chemical formula and, in the main benchmark, the same number of atoms in the unit cell (Omee et al., 12 Jun 2025). ParetoCSP2 is a multi-objective genetic algorithm using energy, age, and space-group diversity control, with adaptive control implemented by minimizing the maximum number of individuals from each space group (Omee et al., 12 Jun 2025). On formulas with two polymorphs and the same number of unit-cell atoms, it achieves a near-perfect average space-group coverage of $f$3 and a complete average StructureMatcher coverage of $f$4; across a regular CSP benchmark it improves over ParetoCSP and GN-OA by $f$5 across the core structure-similarity metrics (Omee et al., 12 Jun 2025). Here, ordered polymorphism is explicitly a multi-solution problem over several symmetry realizations of one stoichiometry.

A complementary materials viewpoint organizes ordered and disordered phases into order-(dis)order family trees. In that framework, a disordered parent is a higher-symmetry average structure with occupational disorder, an ordered child is a lower-symmetry structure derived from that parent, and ordered siblings are different ordered descendants that share the same disordered parent (Yamazaki et al., 23 Apr 2026). The formal basis is group-subgroup descent, especially translationengleiche subgroup relations, with subgroup index

$f$6

The SWORDFamilyMatcher constructs family labels by masking species and recomputing symmetry-aware SWORD labels, and two structures are family-related if

$f$7

The paper reports that for 35 A-Lab target phases, $f$8, meaning 21 trace back to known disordered parents; it also reports that in ICSD about $f$9 of unique ordered compositions show polymorphism, and about $m$0 of those polymorphic systems include at least one pair related by a group-subgroup transition (Yamazaki et al., 23 Apr 2026). In this usage, ordered polymorphism is crystallographic lineage rather than algebraic preservation.

Polymer and supramolecular studies supply two further, related senses of ordered multiplicity. In a coarse-grained polymer melt with two competing incommensurate length scales, spontaneous crystallization does not yield close-packed Fcc/Hcp structures but multiple polymorphs with substantial bond disorder and distorted local packing, which are concluded to be distorted Bcc-like crystals (Giuntoli et al., 2018). The model uses 50 fully flexible linear chains of 10 monomers each, and among 56 initial configurations, 42 crystallized spontaneously while 14 did not within about a month of CPU time; the crystalline states occupy a broad region in $m$1 space, which is the main evidence for polymorphism (Giuntoli et al., 2018). The ordered aspect here is long-range crystalline coherence arising under frustration, not order preservation of an operation.

A related but distinct use of ordering language appears in supramolecular polymers of chemically bidisperse monomers. There the sequence of monomer species along a quasi-linear assembly is modeled by a two-component self-assembled Ising model with

$m$2

so that $m$3 gives blocky order, $m$4 random order, and $m$5 alternating order (Jabbari-Farouji et al., 2010). The paper states that the transition to strong polymerization is described by a critical concentration $m$6 depending on the concentration ratio, and that monomers with a smaller binding free energy are more abundant in short assemblies while monomers with a larger binding affinity are more abundant in longer assemblies (Jabbari-Farouji et al., 2010). This is not a formal theory of ordered polymorphisms, but it shows that in soft-matter contexts “ordering” and “polymorphism” can meet at the level of sequence morphology, fractionation, and coexistence of distinct ordered assembly patterns.

Across these materials literatures, the common theme is multiplicity of ordered realizations under a fixed composition or feed. That usage is structurally analogous to, but formally distinct from, the algebraic meaning of order-preserving polymorphisms.